Home Search Collections Journals About Contact us My IOPscience

Accelerometer based solution for precision livestock farming: geolocation enhancement and

animal activity identification

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2016 IOP Conf. Ser.: Mater. Sci. Eng. 138 012004

(http://iopscience.iop.org/1757-899X/138/1/012004)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 23.95.129.207
This content was downloaded on 05/08/2016 at 13:45

Please note that terms and conditions apply.

II International Congress of Mechanical Engineering and Agricultural Science (CIIMCA 2015) IOP Publishing
IOP Conf. Series: Materials Science and Engineering 138 (2016) 012004 doi:10.1088/1757-899X/138/1/012004

Accelerometer based solution for precision livestock farming:

geolocation enhancement and animal activity identification

G Terrasson1, A Llaria1, A Marra1 and S Voaden1

1
Ecole Supérieure des Technologies Industrielles Avancées (ESTIA),
Technopole Izarbel, 64210 Bidart, France

Email: g.terrasson@estia.fr

Abstract. The rapid evolution of electronics and communication technologies in the last years
has contributed to the expansion of Precision Livestock Farming applications. In this context,
animal geolocation systems applied to extensive farming are interesting for farmers to optimize
their daily work organization. Nevertheless, the deployment of these solutions implies several
technical challenges which must be resolved, mainly the energy consumption and the
suitability of the communication protocols. A recently developed solution that deals with these
technical challenges is the e-Pasto platform, which is composed of low power geolocation
devices embedded into collars that offer an energetic autonomy of at least seven months,
completed with a visualization user interface. The autonomy is assured employing a duty-cycle
operation that results in one geolocation position measurement per hour. This work studies the
employ of accelerometers to overcome this limitation assuring, at the same time, the required
autonomy for the geolocation device. The authors also propose an algorithm that processes the
acceleration data in order to identify the steps of an individual. The whole solution (step
identification and geolocation) has been validated by means of several experimental tests.

1. Introduction
From several years, Precision Livestock Farming (PLF) applications have known an increasing
deployment mainly due to the evolution of the Information and Communication Technologies and to
the miniaturization of the electronic devices. In this frame, the organization of farmers’ daily work
could be optimized by means of tools capable of locating their animals in real time over large area
surfaces. In fact, this kind of applications offers to the farmer the capacity to survey remotely its cattle
avoiding unnecessary displacements and, in addition, to define virtual fences in order to better manage
the pasture resources.
Nowadays, a significant number of geolocation solutions are based on electronic devices,
embedded into collars carried by animals, combined with wireless communication networks that
transmit the collected data. Nevertheless, the application of these systems to extensive farming in high
mountain pastures implies some technical challenges which must be resolved, mainly the energy
consumption and the suitability of the needed communication protocols. In this context, this paper
presents an accelerometer based solution that improves the energy efficiency of embedded geolocation
devices, allowing at the same time a basic animal activity identification.
The document is structured as follows. Section 2 presents the main technical challenges of animal
geolocation in mountain pastures and the e-Pasto platform. The proposed accelerometer based solution
is presented in Section 3, along with the obtained experimental results. Finally, Section 4 concludes
this paper.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
II International Congress of Mechanical Engineering and Agricultural Science (CIIMCA 2015) IOP Publishing
IOP Conf. Series: Materials Science and Engineering 138 (2016) 012004 doi:10.1088/1757-899X/138/1/012004

2. Animal geolocation in mountain extensive farming

2.1. Technical challenges

Geolocation embedded electronic devices, usually carried by the animals, and wireless communication
networks that transmit the collected position data to the final user, are the basic technologies needed to
implement an animal geolocation system. If the geolocation procedure must be performed in large area
mountain pastures, applied to extensive farming, there are a number of technical challenges to be
resolved.
The main constraint is the energy autonomy of the embedded devices, mainly impacted by the
Global Navigation Satellite Systems (GNSS) which are relatively power-hungry. Several methods can
be applied to overcome this problem [1, 2], such as increasing the battery capacity (and thus the size
and weight of the device), energy harvesting technologies (which are not yet very mature) or the
application of a duty-cycle strategy to measure the geolocation positions. This last method obliges to
fix a trade-off to assure both the highest autonomy and the maximum number of collected positions.
Therefore, the main drawbacks of duty-cycle strategy, considering the autonomy constraint, could be
the reduced number of collected position points and the restricted precision.
Concerning the wireless communication networks, the proposed solutions are, in general, based on
the existent wireless technologies, such as GSM, WiFi or ZigBee [3]. Nevertheless, in high mountain
pastures, it is not always possible to guaranty the correct coverage using these technologies. Moreover,
the energy consumption of some of these protocols is excessive for battery powered embedded
electronics, limiting even more the energy autonomy of the geolocation devices.

2.2. The e-Pasto platform

An interesting example of animal geolocation system for extensive farming in mountain pastures,
which deals, at the same time, with the number of measured positions and the battery autonomy of the
devices, is the “e-Pasto” platform [4]. Two main parts compose the e-Pasto solution: the geolocation
devices and the visualization user interface.
The geolocation devices, which are embedded into a collar carried by the animals as illustrated by
figure 1, determine the position via the GNSS module and, afterwards, the collected positions are sent
to a central server by means of the long range low power wireless network developed by SIGFOX©.
As shown in figure 2, the visualization interface allows farmers to see the animals’ measured
positions, as well as other statistical data such as the daily distances covered by each animal. The user
can also define virtual fences in the interface to delimit the authorized zones for their animals. If an
animal goes out of the authorized area, an alarm message is sent to the concerned farmer. This
interface can be accessed from computers or portable devices, such as laptops or smartphones.

Figure 1. Cow carrying the collar Figure 2. Several positions of animals showed in
with the geolocation device. the visualisation user interface.

2
II International Congress of Mechanical Engineering and Agricultural Science (CIIMCA 2015) IOP Publishing
IOP Conf. Series: Materials Science and Engineering 138 (2016) 012004 doi:10.1088/1757-899X/138/1/012004

The correct operation of the overall platform has been validated by means of experimental tests
carried out in mountain pastures of France and Spain in 2013 and 2014. The main enhancement
provided by the e-Pasto platform concerns the autonomy of the embedded electronic devices, which
reaches at least 7 months ensuring, at the same time, the collection of geolocation data in wide surface
pastures, along with an innovative long range and low power wireless communication network.
Nonetheless, this solution is constrained by its duty cycle operation, which limits the animal position
acquisition to one measurement per hour. This duty cycle strategy is very interesting in order to reduce
the overall power consumption of the geolocation devices, but it could decrease the interest of users in
this kind of systems. Certainly, with this strategy it is not possible to collect geolocation data between
two programmed position measurements, even if animals have covered long distances during this
period. In addition, the duty cycle implies a continuous position acquisition, procedure which is not
optimized to ensure an energy efficient geolocation data measurement especially during the night,
when animals stay almost at the same location.

3. Application of accelerometers to improve the geolocation solution

The solution proposed to improve the e-Pasto performances is to incorporate an accelerometer to the
geolocation device. In this manner, geolocation positions measurement is only activated when the
accelerometer detects a precise movement of the animal, procedure which is more efficient in terms of
battery use. The figure 3 compares the principle of duty-cycle operation with the movement detection
one proposed in this article.

Consumption Duty-Cycle Operation Consumption Operation Movement detection

Operation
Operation
Standby
Standby

0 1 2 t (in hours) t
Walking At rest Running

Figure 3. Comparison between duty-cycle and movement detection operation modes.

3.1. Animal activity identification by means of acceleration data analysis

The accelerometer can be employed not only to activate the geolocation measurement, but also to
identify the animal activity and its behaviour by analysing the acceleration data [5]. The table 1
compares different algorithms used for individual activity detection by processing acceleration data.
According to its accuracy to correctly identify different animal movements combined with its low
processing resource requirements and implementation easiness, the most adapted solution for
embedded systems seems to be the moving average method. Consequently, this is the algorithm
adopted for the proposed solution.
Table 1. Comparison of different acceleration data processing algorithms.
Algorithm Advantages Disadvantages
Low resources consumption Low accuracy
Threshold detection
Low energy consumption Impossible to identify the movements
Good movement identification Necessary a high frequency sampling
Frequency analysis
Low error margin High resources and energy consumption
Accurate movement identification Long learning phase
Artificial intelligence
Error margin ≈ 0 High resources and energy consumption
Acceptable resources consumption
Moving average Identification of several movements Not possible to sleep the microcontroller
Easy to implement

3
II International Congress of Mechanical Engineering and Agricultural Science (CIIMCA 2015) IOP Publishing
IOP Conf. Series: Materials Science and Engineering 138 (2016) 012004 doi:10.1088/1757-899X/138/1/012004

3.2. Step identification and geolocation with accelerometers: a practical case

In the context of animal geolocation solution improvement, the first stage is to implement the moving
average algorithm for step identification. In this way, several experimental tests have been conducted
with humans in order to validate the movement identification from the accelerometer data, processed
with the moving average algorithm.
The figure 4 illustrates the principle of the implemented algorithm, and the results obtained for one
axis acceleration data. At t=α, the point A of the curve “average t+2” is lower than point B of the curve
“average t+1”. At the same time, the point B is lower than point C of the curve “average t0”. The
distance between A and C is significant, whereas the distance between A et B is the same as between B
and C. The measure at α validates the “start” of a step. On the other hand, the measure at t=δ validates
the “end” of a step. At t=β, the distance between A and C is not significant enough, and this measure
has not been taken into account, which is also the case for t=γ. This algorithm allows the step
detection and, in addition, it can identify, from the distance between the points A and C, if the
individual is walking or running.

Figure 4. Acceleration curves employed for step detection.

A program that implements the proposed algorithm has been developed and tested with the
TD 1204© evaluation electronic card, which is composed of the same elements than the e-Pasto
geolocation devices: a low-power microcontroller, a GPS module, and a SIGFOX© modem. In
addition to these elements, the TD 1204© also includes an accelerometer, which was not present in the
e-Pasto devices. During the experimental tests, the data issued from the accelerometer were processed
by means of the proposed algorithm to identify the steps. The GPS module was activated every 50
steps, and the measured geolocation position was transmitted to a data base through the SIGFOX©
wireless network. The figure 5 shows in green the geolocation points measured every 50 steps during
the tests. The real path, obtained with a classical GPS device of a smartphone, is shown in red to prove
the precision of the developed solution. Finally, the figure 6 shows the block diagram of the e-Pasto
geolocation device, including the accelerometer.

Figure 5. Geolocation positions Figure 6. Block diagram of the e-Pasto geolocation device
obtained every fifty steps. including an accelerometer.

4
II International Congress of Mechanical Engineering and Agricultural Science (CIIMCA 2015) IOP Publishing
IOP Conf. Series: Materials Science and Engineering 138 (2016) 012004 doi:10.1088/1757-899X/138/1/012004

4. Conclusion
This paper has presented a solution, based on accelerometers, that is able to identify the steps of an
individual. This technique is useful to collect animal behaviour information and also to use in a more
efficiently manner the energy of geolocation devices dedicated to PLF applications. Nevertheless,
several aspects must be discussed.
Concerning the impact over the autonomy, the proposed solution improves the energy efficiency
because of the optimization of the geolocation positions collection procedure: the GPS module is only
activated if a significant movement of the animal is detected by the accelerometer. In addition, the
energy consumption due to accelerometers is not very high, and the computational load of the moving
average algorithm is moderate, because it is applied to only one acceleration axis. However, the
impact of the energy consumption due to both the accelerometer and the processing algorithm over the
geolocation device autonomy must be analysed more in detail and, specially, in a high mountain
pasture scenario.
The proposed solution has been validated with humans and it is necessary to study its application to
animals. The correct positioning of the electronic device over the animal is crucial to achieve the step
detection. Related to the interpretation of the acceleration curves, the scientific literature [5] shows
that the results obtained for different kinds of animals are similar to those obtained for humans.
Finally, it would be desirable to detect, with an adapted algorithm, other kind of animal movements
and activities (eating, rumination…), respecting the trade-off with the autonomy. This type of
information could be used by the farmers in order to determine the well-being of their cattle.

5. References
[1] Ruiz-Mirazo J, Bishop-Hurley G and Swain D 2011 Automated animal control: can
discontinuous monitoring and aversive stimulation modify cattle grazing behavior?
Rangeland Ecology Management 64 240–8
[2] Gutierrez A, Dopico N, Gonzalez C, Zazo S, Jiménez-Leube J and Raos I 2013 Cattle-powered
node experience in a heterogeneous network for localization of herds IEEE Trans. Ind.
Electronics 60 3176–84
[3] Raychaudhuri D and Mandayam N 2012 Frontiers of wireless and mobile communications
Proc. IEEE 100 824–40
[4] Llaria A, Terrasson G, Arregui H and Hacala A 2015 Geolocation and monitoring platform for
extensive farming in mountain pastures Proc. IEEE Int. Conf. on Industrial Technology
(Seville) pp 2420–5
[5] Grundy E, Jones M, Laramee R, Wilson R and Shepard E 2009 Visualisation of sensor data
from animal movement Proc. IEEE-VGTC Symposium on Visualization (Berlin) pp 1–8

Acknowledgements
The work presented in this paper has been developed in the frame of the e-Pasto project (pilot project
of AGRIPIR), which has been financially supported by the European Commission in the frame of the
Spain-France-Andorra Territorial Cooperation Operational Program (POCTEFA 2007-2013) and the
FEDER funds.